propagation of microwaves in the troposphere with potential...
TRANSCRIPT
chaytr'r 3 71
~haptrr 3
D£V£LOPMENT Of TROPOSPH£RIC Dt:LA Y MODELS
3.0 Introduction
Neutral atmosphere being non-dispersive in nature, the only way to mitigate the
tropospheric delay is to estimate it from the refractivity profiles, accomplishing a priori
knowledge of which to incorporate in operational model is rather difficult. Then it becomes
essential to develop conventional models relating the Tropospheric delay (both ZHD and
ZWD) with easily measurable atmospheric parameters at the surface. Though a typical mean
value of - 2.3 m for ZHD and 20 to 40 cm for ZWD can generally be taken as a "rule of
thumb", the day-to-day variability in atmospheric parameters can introduce a variation up to
±15 cm about this mean value, as far as tropospheric range error is concerned. Because of
this, conventional models [Saastamoinen, 1972; Davis et al., 1985; lfadis, 1986; Mendes
and Langley, 1998] for ZHD and ZWD are developed in terms of atmospheric parameters at
the surface, such as surface pressure (Ps), surface temperature (Ts) and surface water vapor
pressure (es).
Saastamoinen [1972] developed a simple relation between ZHD and Ps for different
latitudes taking into account the altitude variation also, which is fairly accurate up to 0.4 km
for different seasons. Later Davis et al. [1985} and Elgered [1991] proposed similar models
for ZHD with a slightly different value for the relevant coefficients. All these models are
developed under the assumption that the dry atmosphere is in hydrostatic equilibrium and
follows the equation of state. However, since water vapor need not be in hydrostatic
equilibrium and is heterogeneously distributed in the atmosphere, similar models were not
attempted for ZWD until lfadis [1986J who developed simple empirical relation between
ZWD and ps. Ts and es. Later Mendes and Langley [1998] also proposed a simple empirical
relation for ZWD using es and Ts. A slightly different model for ZHD and ZWD are evolved
by Hopfield [1971], following a different approach based on refractivity profile.
A brief outline of these models is summarized in Chapter 1 Section 1.8. It would be
worth in this context to note that all these empirical models are developed based on
meteorological data from European and North American continents [Satirapod and
c'hayter 3 72
Chalermwattanachai, 2005]. The accuracies of these models, assessed by comparing with
the true estimates obtained from ray-tracing, are typically < 3 mm for ZHD [Janes et al.
1991], and < 5 cm for ZWD [Askne and Nordius, 1987; lanes et al., 1991; Elgered, 1992]
with a maximum deviation extending up to - 8 cm during abnormal weather events
[Elgered, 1993; lchikawa, 1995]. However, no such studies are reported till date over the
tropical/subtropical region. Taking note of this fact, empirical models applicable for the
tropical region are developed based on the meteorological data from different locations
situated at distinct climatic conditions over the Indian subcontinent. These models are
validated by comparing with the true range error (estimated by ray-tracing).
3.1 Site-Specific Surface Models for Indian Subcontinent
The relationship between zenith tropospheric delay and the atmospheric parameters at
surface are examined for different stations located at different climatic zones over the Indian
subcontinent. A simple linear relationship is established for ZHD in terms of surface
pressure through regression analysis. The true value of ZHD is estimated by applying the
integral (eg. 2.39) to the monthly mean atmospheric models described in Section 2.6 and an
empirical linear model connecting ZHD with Ps is developed in the following form
ZHD = aD' Ps (3.1)
Figure 3.1 shows typical scatter plot of ZHD estimated through ray tracing with Ps (in hPa)
for eight among the eighteen stations selected for this analysis (Table 2.1). These stations
represent the major climatic zones over the Indian region and also have the location in
proximity to major airports, considered in the GAGAN project. These stations are grossly
referred to as the Range Integrity and Monitoring (RIM) stations. The regression analysis is
performed keeping the value of intercept as zero and the resulting slope (aD) and correlation
coefficient are listed in the respective scatter plots. For all these stations, the value of
correlation coefficient lies between 0.82 and 0.99. The standard error of aD for each station
is < 0.024% which corresponds to about 0.05 cm in delay. The standard deviation (SD) of
the correlation also is listed in these plots. The values of coefficients aD thus estimated for
all the stations are presented in column 2 of Table 3.1.
Following a similar approach linear models are established for ZWD on a monthly
mean basis for different stations with different surface humidity parameters such as surface
--e ->,
= -~ Q ~ '. -= -rIl 0 1-0
"C >,
== ..c -.• I:: ~
N
2.32
2.31
2.30
2.29
2.28
2.27
"0.0.002281:3.3"10"
R = 0.82 SO = 0.002
la) TRIVANORUM
999 1000 1001 1002 1003 1004 1005
2.11.--------------------, 110 = Q.00228 :!:3.2'10·'
2.10 R=0.97; SO = O.0014m
2.09
2.08
2.07
2.06
Ic) 2.05
904 906 908
2.32 "0 = 0.00229 .! 4.4'10"
2.31 R = 0.98 SO = 0.00216
2.30
• 2.29 ••
• 2.28 • •
2.27
(e) 2.26
992 996 1000
2.27 a, = 0.00%3 .! 4.73'10"
2.26 R = 0.99 SO = 0.00226
2.25
2.24 •• •
2.23 •
2.22
Ig) 2.21
964 988 972
BANGALORE
910 912 914
• ••
•
•
AHMEDABAD
1004 1008 1012
•• •
DELHI
978 980 984 968
2.34 110 = 0.00231 .! 5.4'10"
2.33 R = 0.87 SO = 0.00381
2.32
2.31
• 2.30
2.29
(b)
•
PORTBLAIR 2.28 +--"T""~_r~--,---<'-.-___,--r__~r__,.......j
996 997 998 999 1000 1ilO1 1002 1ilOJ 1004
2.34,------------------.," 0.00229:!: 3.5'10·'
2.33 R " 0.99; SO = 0.0017
• •• 2.32
• 2.31 • •
2.30
2.29 • 2.28
Id) KOLKATA
996 1000 1ilO4 1008 1012 1016
Surface Pressure (hPa)
73
Figure 3.1: Scatter plots along with the regression line showing the variation of ZHD with Ps for different stations located at different climate zones over the Indian subcontinent and adjoining oceanic region
chayt(f 3 74
Table 3.1: Coefficients of Site-Specific Surface Models for Different Indian Stations
Dry Wet
Station ZHD (m) = aJJxPS ZWD (m) = aw><es ZWD (m) = bw'Xps ZWD (m) = CwX PW
aDxlO·3 aw bw Cw xlO·3
Trivandrum 2.280 ± 3.3E-4 0.01199 ± 1.5E-4 16.595 ± 0.233 6.45 ± 5.6E-3 (0.82; 0.002) (0.77; 0.0217) (0.72; 0.023) (0.99; 0.0015)
Portblair 2.310 ± 5.4E-4 0.01212 ± 1.8E-4 16.753 ± 0.255 6.45 ± 4.3E-3 (0.87; 0.0038) (0.93; 0.025) (0.91; 0.025) (0.99; 0.0011)
Bangalore 2.21\0 ± 3.2E-4 0.0122 ± 2.48E-4 16.703 ± 0.384 6.53±7.IE-3 (0.97: 0.0(14) (0.86; 0.025) (0.8 I; 0.0278) (0.99; 0.0013)
Mangalol"c 2.280 ± 3.4E-4 0.0113 ± 2.2E-4 15.628± 0.316 6.35 ± 7.7E-3 (0.89; D.OO 17) (0.85; 0.030) (0.85; 0.032) (0.99; 0.00) 9)
Chcnnai 2.290 ± 2.7E-4 O.O[ 11 ± 3. lE-4 15.409± 0.443 6.33 ± 7.8E-3 (0.98; 0.0013) (0.58; 0.042) (0.54; 0.043) (0.99; 0.0019)
Goa 2.280 ± 6.6E-4 0.0[09± 2.9E-4 15.047± 0.40 6.34±7.1E-3 (0.81; 0.0(32) (0.92; 0.040) (0.92; 0.040) (0.99; 0.0017)
Hyderabad 2.230 ± 3.0E-4 0.0142 ± 2.5E-4 19.577± 0.413 6.42 ± 5.6E-3 (0.98; 0.0(14) (0.94; 0.026) (0.90; 0.(30) (0.99; 0.0013)
Vishakhapatnam 2.270 ± 3.4E-4 0.QI07± 2.8E-4 15.477± 0.87 6.39 ± 7.8E-3 (0.99: 0.0017) (0.81; 0.039) (0.41; 0.089) (0.98; 0.0200)
Mumbai 2.290 ± 3.7E-4 0.0117± 3.8E-4 16.267± 0.56 6.33 ± 5.5E-3 (0.97; 0.(017) (0.92; 0.049) (0.90; 0.051) (0.99; 0.0013)
Kolkata 2.290 ± 3.5E-4 0.0119 ± 3.01E-4 16.43 ± 0.427 6.43 ± 9.2E-3 (0.99; 0.0(17) (0.94; 0.0408) (0.94; 0.0418) (0.99; 0.0023)
Ahmedabad 2.290 ± 4.4E-4 0.0119 ± 2.27E-4 16.627 ± 0.315 6.43 ± l.IE-2 (0.98; 0.0022) (0.97; 0.0245) (0.97; 0.0243) (0.99; 0.0022)
Bhopal 2.160 ± 6.IE-4 0.0112± 2.5E-4 15.394± 0.38 6.37 ± 3.7E-3 (0.97; 0.0(28) (0.96; 0.023) (0.95; 0.025) (0.99; 0.0060)
Guwahati 2.290 ± 3.0E-4 0.0[27 ± 2.7E-4 17.520 ± 0.399 6.46 ± I.IE-2 (0.99: 0.0(15) (0.97; 0.0338) (0.97; 0.(36) (0.99; 0.0027)
Jodhpur 2.230 ± 7.1 E-4 0.0122± 2.0E-4 16.847± 0.30 6.34 ± 0.2E-3 (0.98: 0.0(32) (0.98; 0.018) (0.98; 0.019) (0.99; 0.0028)
Lucknow 2.260 ± 5.2E-4 0.0 126± 4.5E-4 17.418±0.66 6.25 ± 0.3E-3 (0.98: 0.0025) (0.95; 0.048) (0.95; 0.051) (0.99; 0.0069)
Delhi 2.300 ± 4.7E-4 0.0128 ± 2.3E-4 17.539 ± 0.373 6.44 ± [.4E-2 (0.99: 0.0(23) (0.99; 0.023) (0.98; 0.0274) (0.99; 0.0024)
Patiala 2.200 ± 9.4E-4 0.0101± 8.4E-4 13.929± 1.15 6.00 ± 1.13E-1 (0.95; 0.0044) (0.70; 0.084) (0.70; 0.083) (0.99; 0.0069)
Srinagar 2.280 ± 6.5E-4 0.0118 ± 1.5E-4 15.934±0.[72 6.61 ± 2.5E-2 (0.97; 0.0(27) (0.99; 0.0093) (0.99; 0.0078) (0.99; 0.0027)
75
water vapor pressure (es in hPa), surface water vapor density (Ps, in kg m-3) and the
columnar integrated water vapor (Is in kg m-2) also known as Precipitable Water vapor (PW
in mm). As a typical example the scatter plots of ZWD estimated through ray tracing, with
es for the eight RIM stations are presented in Figure 3.2. A linear relationship was sought
analogous to that of the dry component through regression analysis keeping the value of
intercept as zero in the following form
ZWD = aw x es
ZWD =bw xPs
ZWD=cwxPW (3.2)
values of the coefficients thus estimated are presented in columns 3,4 and 5 of Table 3.1 for
all the eighteen stations. The quantities shown inside the parenthesis in each cell are the
correlation coefficient and its standard deviation, respectively. A perusal of Figure 3.1 along
with Figure 3.2 shows that the deviation of points from the mean line is relatively large in
the latter case, the effect of which is also observable in the standard error associated with the
regression coefficients. This shows that, in general, the uncertainty in the prediction of ZWD
is larger than that of ZHD. The regression coefficients connecting ZWD with es, Ps and PW
are also examined by removing the imposed condition of "zero intercept". The values of
regression coefficient are rearranged in a more convenient form as shown below and the
coefficients ao, bo, Co, eo, PG, and PW u are estimated.
ZWD = ao x (es - eo)
ZWD = ho x (Ps - Po)
ZWD = Co x (PW - PWo)
for es > eo
for Ps> Po
forPW > PWo (3.3)
The values of these coefficients along with their respective standard error are summarized in
Table 3.2. The respective values of the correlation coefficient and its standard deviation are
presented within the parenthesis just below the corresponding coefficients. The values of ao,
and bo are marginally greater than corresponding values of aw, and bw for all the 18 stations
except Chennai, Hyderabad, Vishakhapatnam, Bhopal, Jodhpur, Patiala and Srinagar where
the values of es, Ps and PW are relatively small. The values of Co and Cw for different
stations are more or less comparable. As these models are developed for each station
separately based on the monthly mean atmospheric models and surface meteorological
parameters from respective stations, they are termed as "Site-Specific surface models" in
chaytir 3
0.6,------------------, la) TRIVANDRUM
0.4
0.3
0.2
0.1
25 26 27 28 29 30 31
0.5,------------------,
0.4
0.3
0.2
~ 0.1 ~ ..-~
(c) BANGALORE
1"'\ 0.0 +-~~_._~----r~---.-~..__~_.__I ~ 14 16 1B 20 22 24 ..... ~ > O.S
;;> (e)
-s 0.4 ..... I=l ~ N 0.3
0.2
0.1
AHMEDABAD
•• • -- ..
-
0.0+--~--r-~__,.-~-r_--,_~-1
10 15 20 25 30 35
0.5,-------------------, DELHI Ig)
•• 0.4
0.3
0.2 - _.
0.1
0.5,.....---------------~ (b) PORTBLAIR
0.4
0.3
0.2
0.1
0.0 +-~.-~.--.__,.-~-_.__~""T"'--r-'T"""~._,.......j 23 24 25 28 27 28 29 30 31 32
0.5..-----------------~
Cd)
0.4
, 0.3
• • • • • 0.2 -- -0.1
KOLKATA
15 20 25 30 35 40
0.5.,.------------------~
(11
0.4
• 0.3
0.2
0.1
•
-
--,,-
GUWAHATI 0.0 +-_r----r-_-..,-_,._-...--..,--,--~-
10 15 20 25 30 35
0.5,.....---------------~
(hI SRINAGAR
0.4
0.3
0.2
0.1
0.0 0.0 +-...-..---,_-..-~__r_~T""""~ _ _,._~...__.........,,._,.___l
5 10 15 20 25 30 35 4 6 8 10 12 14 16 18 20 22
76
Surface water vapor pressure (hPa) Figure 3.2: Scatter plots along with the regression line showing the variation of ZWD with es for different stations located at different climate zones over the Indian subcontinent and adjoining oceanic region
chaytr'r 3 77
Table 3.2: The Regression Coefficients Connecting ZWD with eST Ps and PW, Removing
the Imposed Condition of "Zero Intercept"
Wet range error
ZWD (m) ;;; ao X (es-eo) ZWD (m) ;:: bo X (Ps-Po) ZWD (m) ;:: Co X (PW- PWo) Station
(ao±Aao) eo±Mo boiMo
(Pr,)±f1f}u) (co±&:o) PWu±f1PWo xlO-3 X 10.3 X 10.3
Trivandrum 15.11 ± 2.7
7.82 ± 0.08 21.39 ± 4.5
5.98 ± 4.5 6.35 ± 0.05
0.19±2e-3 (0.77;(1.019) (0.71 ;0.02) (0.99;0.001 )
Portblair 19.17±1.5
10.59 ± 0.04 24.47 ± 2.3
6.58 ± 2.3 6.46 ± 0.03
0.07 ± 1.6e-3 (0.93;0.018) (0.91;0.021) (0.99;0.00 I)
Bangalore 14.13 ± 1.8
2.80 ± 0.04 17.65 ± 2.7
0.806 ± 2.7 6.56 ± 0.04
O. [74 ± 1.4e-3 (0.86;0.024 ) (0.81 ;0.028) (0.99:0.00 I)
Mangalore 17.34 ± 2.3
9.9 ± 0.06 25.69 ± 3.4
8.03 ± 3.4 6.47 ± 0.04
1.00 + 2.2e-3 (0.85;0.027) (0.85;0.027) (0.99;0.002)
Chcnnai 11.95 ± 3.6
2.07 ± 0.10 15.99 ± 5.4
0.73 ± 5.4 6.30 ± 0.05
0.27 ± 2.3e-3 (0.58;0.043) (0.54;0.044) (O.99;0.002)
Goa 18.81 ± 1.7
11.89 ± 0.05 25.69 ± 2.4
8.5 ± 2.35 6.42 ± 0.03
0.567 ± 1.3e-3 (0.92;0.030) (0.92;0.030) (0.99;0.002)
Hyderabad 15.53 ± 1.2
[.77 ± 0.03 21.22 ± 2.1
1.18 ± 2.1 6.40 + 0.02
0.[5 ± l.le-3 (0.94;0.026) (0.9! ;0.030) (0.99;0.00 I)
Vishakhapatnam 11.45 ± 1.8
1.95 ± 0.05 12.3 ± 5.94
5.43 ± 5.9 6.85 ± 0.30
3.6 ± J.5e-2 (0.81 ;0.039) (0.40;0.090) (0.9R;0.OJ9)
Mumbai [6.71 ± 1.6
8.14 ± 0.04 23.10 ± 2.4
5.78 ± 2.4 6.33 ± 0.02
0.07 ± 1.0e-3 (0.92;0.041 ) (0.91 ;0.044) (0.99;0.00 I)
Kolkata 13.80 ± 1.0
3.92 ± 0.03 [9.81 ± 1.5
3.53 ± 1.5 6.34 ± 0.02
0.79 ± 1.0e-3 (0.94;0.038) (0.94;0.038) (0.99;0.002)
Ahmcdabad 12.37 ± 0.6
0.86 ± 0.01 16.98 ± 0.9
0.35 ± 0.8 6.37 ± 0.03
0.45 ± l.le-3 (0.97;0.025) (0.97:0.U25) (0.99:0.002)
BhopaJ 11.08 ± 0.7
0.13±0.01 15.16±1.1
0.22 ± 1.0 6.13±0.08
1.42 ± 2.7e-3 (0.96;0.023 ) (0.95;0.Q25) (0.99;O.005)
Guwahati 15.69 ± 0.8
4.96 ± 0.02 22.44 ± 1.5
4.19 ± 1.2 6.34 ± 0.02
1.05 ± 0.8e-3 (0.97;0.027) (0.97 ;0.027) (0.99;0.00 I)
Jodhpur [2.12 ± 0.5
0.04 ± 0.01 17.42 ± 0.8
0.50 ± 0.7 6.22 ± 0.Q2
0.82 ± 0.ge-3 (0.98;0.018) (0.98;0.019) (0.99;0.002)
Lucknow [5.59 ± 1.0
4.59 ± 0.Q2 21.96 ± 1.6
3.60 ± 1.6 6.05 ± 0.05
1.72 ± 2. le-3 (0.95;0.041 ) (0.95:0.044) (0.99:0.005)
Delhi 14.30 ± 0.5
2.34 ± 0.01 20.20 ± 0.7
2.14±0.8 6.34 ± 0.01
0.76 ± 0.7e-3 (0.98;0.019) (0.98;0.022) (0.99;O.002)
Patiala 9.90 ± 2.2
0.43 ± 0.05 14.11 ± 3.2
0.21 ±3.2 5.50 ± 0.20
4.15 ± 6.2e-3 (0.70;0.086) (0.70;0.085) (0.99;0.016 )
Srinagar 11.70 ± 0.4
0.15 ± 0.01 16.58 ± 0.4
0.39 ± 0.4 6.39 ± 0.03
0.87 ± 0.6e-3 (0.98;(J.OO9) (0.99;0.007) (0.99;0.001)
78
further discussions. However, the accuracy of these models in general depends on the
accuracy of the surface atmospheric parameters [Back and Doerflinger, 2000].
3.2 Adaptation of Hopfield Model for the Indian Region
The surface models provide integrated range error all the way from top of the
atmosphere to the surface. Whenever the prediction of tropospheric range error is required
from an elevated platform, a priori knowledge of the altitude dependences of the relevant
coefficients is to be accounted through an appropriate correction factor [Baby et al., 1988].
Taking this factor into account, the Hopfield [1969] model (which addressed the altitude
dependence in range error) is adapted for the Indian region. As detailed in Section 1.5.4, in
this model the integral effect of refractive index parameter above the user altitude alone is
considered. The altitude variation of refractivity is modeled through the semi-analytical
relations connecting the surface refractivity with the characteristic height parameter.
Correspondingly two characteristic height parameters evolve, one for the hydrostatic
component (hD) and the other for the non-hydrostatic component (hw). The altitude profile
of dry component of refractivity (ND) when h $ ho can be expressed as,
(3.4)
where ho (in km) is the "characteristic height" for dry refractivity profile, hs is the altitude of
the station above MSL and Nso is the value of ND at the surface (where h = hs). The equation
for dry component is obtained by analytically solving the gas law equations applicable for
the atmosphere [Hop field, 1969]. As the temperature lapse rate varies with geographic
latitude as well as to certain extent with altitude also, " also varies accordingly. Even
though, no such analytical form can be arrived in the case of the wet component (Nw),
Hapfield [1969] in his formulation assumed that the water vapor also follows a similar law
with same value of "but different value of characteristic height (hw). With this assumption
the altitude variation of wet component of refractivity (Nw), when h $ hw, is modeled as
(3.5)
where hw (in km) is the 'characteristic height' for wet refractivity profile and Nsw is the
value of Nw at the surface. In the above formulations, it is implicitly assumed that the effect
79
of ND above hD and that of Nw above hw is insignificant. As the zenith delay is obtained by
integrating the atmospheric refractivity along the altitude between hs and top of the
atmosphere, ZHD and ZWD can be estimated by analytically integrating eg. (3.4) and eq.
(3.5) respectively, replacing the upper limits by the respective characteristic heights. If we
consider the altitude profiles of ND and Nw starting from the surface, for any location hs can
be treated as zero. The expression for zenith range errors can then be written as.
hIJ h ZHD = fNDdh = NSD __ D_
o (17+ 1) (3.6a)
and
hw hw ZWD= fN .... eldh=Nsw--
o (17 + 1) (3.6b)
The above equations are strictly valid only when the integrals are taken up to the respective
characteristics height.
If we consider the mean sea level values of the dry and wet components of the
refractivity (NOD and Now, respectively), then NSD and Nsw at the surface (h = hs) can be
written in terms of their MSL values as
(3.7a)
and
[hw -hs)1I
Nsw =Now hw (3.7b)
Then eq. (3.6a) and eg. (3.6b) can be rewritten accordingly in terms of NOD and Now.
Following a similar approach, eq. (3.7a) and eq. (3.7b) both can be extrapolated for any
given altitude by replacing hs with that altitude. In case if one is interested in using the
surface values NSD and Nsw, same procedure can be followed by replacing NOD and Now in
eq. (3.7a) and eq. (3.7b), respectively, by NSD and Nsw. In this case hs should be replaced by
the height of the location above the surface (hp).
However, this method requires a priori knowledge of the characteristic height
parameters and atmospheric refractive index at the surface. The characteristic height is
derived by equating the numerical integral of refractive index profile to the theoretically
estimated value of integrated refraction effect given in eq. (3.6a) and eq. (3.6b). In this
80
process f3 decides the value of 7] and hence the denominator of the eq. (3.6). Even though
using different temperature lapse rate will change the value of 7], the value of hD estimated
through eq. (3.6a) using the new lapse rate also will change correspondingly in such a way
that the ratio of ho/( TJ+ 1) remains the same. This is true in case of hw also. However, in the
present study the temperature lapse rate is taken as 6.7 Klkm as was done by Hopfield
[1971] and tre values of hD and hw derived for all the stations are based "n this lapse rate.
Implementation of this model, however, involves modeling of the characteristic height
parameters for different stations. Substituting the true range error estimated by integrating
the refractivity profiles obtained from the monthly mean atmospheric models in eq. (3.6a)
and eq. (3.6b) respectively; the dry and wet characteristic heights are estimated at each
station for different months. The month-to-month variation of ho estimated for 00:00 UTC
and 12:00 UTe (corresponding to 05:30 1ST and 17:30 1ST, respectively) for all the
eighteen stations used for the present study are presented in Figure 3.3(a-r). Similarly,
Figure 3.4(a-r) shows the month-to-month variation of hw for these stations. The vertical
lines in these figures show the respective standard deviations in each month (representing
the day-to-day variability) estimated by applying the error propagation fonnula [Ku, 1966]
appropriately. While the month-to-month variation (seasonal) of ho is small for the tropical
stations « ±0.5 km), it is large for the high-latitude stations (- 5 km). While the mean value
of hw varies between - 11 km to - 14 km for the tropical stations it varies in the range 8 km
to 16 km for the high latitude stations. Significant annual variation in hw is observable for
the high-latitude stations (with amplitude as large as - 9 km). A distinct morning to evening
difference is observable in the values of ho for all these stations with the evening values
consistently being larger than the corresponding morning values, even though such a trend
does not emerge for hw. However, in most of the cases the evening values of hw are larger
than the morning values. The annual mean values of ho and hw for the eighteen stations are
presented in Table 3.3 along with the respective standard deviations, within parenthesis
(indicating the extent of mean monthly variations). While the observed mean values of ho
and hw are quite large and fairly constant for the near equatorial stations (Trivandrum,
Portblair and Bangalore), a prominent annual variation is observable for the other stations
particularly in case of ho. The characteristic height for different months can be used as a
look up table for each station for estimating the range error employing eq. 3.6(a) and eq.
3.6(b) by incorporating the respective values of surface refractivity.
chaytt'T 3 81
Table 3.3: Mean Value of hD and hw along with the Maximum Month-to-month Variability (within parenthesis)
Stations hD (km) hw (km)
Trivandrum 44.610 (0.39) 13.32 (0.88)
Portblair 44.580 (0.25) 13.34 (1.10)
Bangalore 43.890 (0.70) 13.28 (1.59)
Mangalore 44.270 (0.43) 12.69 (1.16)
Chennai 44.363 (0.52) 12.61 (1.81)
Goa 44.258 (0.46) 12.06 (1.67)
Hyderabad 43.098 (0.75) 15.99 (1.93)
Vishakhapatnam 43.904 (0.39) 12.96 (3.90)
Mumbai 44.278 (0.58) 12.79 (2.12)
Kolkata 44.063 (0.79) 12.83 (1.60)
Ahmedabad 44.323 (1.30) 13.25 (1.46)
Bhopal 41.509 (1.02) 12.76 (2.22)
Guwahati 43.750 (0.77) 13.41 (1.71)
Jodhpur 42.814 (1.19) 14.02 (1.41)
Lucknow 43.282 (1.08) 13.29 (3.04)
Delhi 43.910 (1.17) 13.34 (1.80)
Patiala 41.673 (1.15) 12.14 (9.92)
Srinagar 41.9600.32) 12.08 (0.95)
- E .:.:: - -.t:. C) .- Cl) .t
:.
U .- -tI) .~ C
l) -u re: ~
re: .t:.
o ~
c
47~----------------------------------,
46
45
44
43 ::J
TR
IVA
ND
RU
M
12
00
UT
e
_ ...
. D--
o)-
-" __
__ ~
T
_~
o~-~
I-"~--;-Q-l~-~
... _.-_
_ --"
! .. ---.~.--~~-.-.-.
-.-6
'~_ ...
0
00
0 U
Te
, I
lill
, ,
, ,
, I
I I
• I
, ,
2 3
4. 5
6 7
8 9
10
11
12
47 J M
AN
GA
LO
RE
46-1
12
00
ure
4
5
44
'.:~ .
. : · .. :>
~:~t~-
=l: ...
-... -~
. 00
00 U
TC
.
", ±
• ~ ,_
,.:
J.
~,_"
__
~r'
43
42
Idl
41~1-r~~~~'-~~-r,-~-.~.-~-r-r~~
23
45
6
8 9
10
11
1
2
47
,-----------------------------------,
46
45
44
43
42
HY
DE
RA
BA
D
12
00
UT
e
l·-l/
··''--
..l
1,/
/1
I"
" ",
r/
.. ; ..
.. 11
,l/"
l "'.
,t--·1
~ -.:
; -~.
),/o~oo
UT
e
.!! '''.. "
.. ~
/'~
1.·
A
,· .. 1
,
I (0
) 41~1_.~~,_-.~~_.~~,_~~~_r~r4
.1,-
----
----
----
----
----
----
----
----
----
, P
OR
TB
LA
IR
48
12
00
UT
C
~ _
__
g _
__
c--
J _
__
,~. -
-i--'" _
I' ~_
"'l
~ • .
...---
A-&.
..-"-"
_~.
..-.
..i.
~,!.
-_.
--:l.
...-_
.Ii--
r: A
_ ..
__
_ ,
'!IP~~-.i.-
...
46
44
0000
UT
C
43
::l , I
(b)
, ,
, ,
, ,
i ,
, ,
, 1
2 3
4 5
6 7
8 9
10
11
1
2
47
J CH
EN
NA
I
46
12
00
UT
C'
45
..-;
..,.0'1
"'_
,1 .
44
."
,A"
,.,,
& .. -
~ -.
-!.
-.~~
...•
_".~
",,-,
~:.
1/
00
00
UT
e
'''"
. .1.,
. "
-, .. ~
...I. •
. '
",./
"
43
42
47
,--------------------------------------,
BA
NG
AL
OR
E
-46
T_--~.-~
;;;--
--.'"
1~
T,
12
00
UT
C
,,/
r'--
-~_
;:; -2. -
_~ __
: __
_ ):
.l.
45
44
.A-a.
l---
-~ --.-+_+_~_.~
/.1
0
00
0 U
TC
'--
-I
!~
~ 43
42
41 I ,
(c
) , I
, ,
, ,
1 2
3 4
5 6
1
GO
A
47
F
46
45
1
20
0 U
Te
....
:"-
. ~ ....
"'" ~J
!. '
.~~ .. o
,'" ~-
-b'''
~ "~
<!~
44
43
42
/-
00
00
UT
e
..&
--~ .......
.
8 11
10
11
1
2
l.
. . ~
..:.
, •• :
;. -
_.
1
" ..
....
....
....
....
. 4 __
~.~_, +
10
1
1
12
4
1 i,
, (f
) 4
7
• 2
i ~
5 ~
7 ~ M
UM
BA
I 8
9 1
0
11
12
(e)
41~1-r~~,_~~~_r~r_,_~~~_r~r4
2 4
5 7
8
47,-------------~------------------__,
VIS
HA
KH
AP
AT
NA
M
46
46
, .. ~2
00 U:.:.
~~:::·.t
.-*~·:=~
-:-~ .-
L~:~~.
,~.~
._A
' ~~~o
ure
-"'
t
45
44
43
45
44
12
00
UT
e
:.--I
·-:
:' ·-::
:.L::L
_ .. /,
-.
---.. _-
. .I
. ,/
'·~L.I
1/'.1
.~
1 l
._.-
( 0
00
0 U
Te
,,,,.,
.. ..... ~.
~ '-
43
42
42
.') 41
I
.h)
. .
41
l
i I
i'
i i
I'
i i
i r
, I
I 4
5 6
T
8 9
10
11
12
3
4 5
• T
8
9 10
11
1
2
1 2
3 4
5 8
7 8
9 1
0
11
1
2
Mon
th O
f Y
ear
Mon
th O
f Yea
r M
onth
Of
Yea
r
Fig
ure
3.3
(a-i
): M
on
th-t
o-m
on
th v
ari
atio
n o
f ho
fo
r T
riva
nd
rum
, P
ort
bla
ir,
Ba
ng
alo
re,
Ma
ng
alo
re,
Ch
en
na
i, G
oa,
Hyd
era
ba
d,
Vis
ha
kha
pa
tna
m,
an
d M
um
ba
i. co
nti
nu
ou
s lin
es
wit
h s
olid
tri
an
gle
re
pre
sen
t th
e v
ari
ati
on
s fo
r 0
0:0
0 U
TC a
nd
th
e s
am
e w
ith
o
pe
n s
qu
are
fo
r 1
2:0
0 U
TC
«.
;:r-
-,
~ 3-~ 00
N
- E .:.:: - -.s::. 0)
'CL>
.s::.
(,J .- .... en
.t:
CV -(,J n:J
lo ns
.c
U ~
C
41
K
OL
KA
TA
46
..j
45-1
T
... _
t··,
·r·"
,,1
12
00
ure
).,
1 ').
.. -.J
: :t.
/1
1
l',,-
"r--
1-
·Co
~-~-il-._ I~
44-1
T
/1
0
~.
1"-,
,,
'" '/
' ... ~ ,
42
-1 t
1 '1
41
I
(J)
2 3
4 5
6 8
9 1
0
11
1
2
47
,-----------------------------------4
G
UW
AH
AT
l
46
45
12
00
UT
e
I 1
_
{---I
--~ .. -,
',_T,
)~
·'r-
~/~ ..
'~;U·TC--~~~~
.... y
J/
. . "'~;'
r:/
)/'
I'\.
43
-1
., /
"-1
1 /
, I--
-'±
42
-1
J.
44
41
I (r
n)
• i
4 5
6 7
8 9
10
1
1
12
47
'DE
LH
I
, A ... 1.
i,
/ j'''
-.'.
T 1
20
0 U
Te
//
~ 1:
"-,,,,~.~
,.' i",)
I!
)/]'
-~.-..
~~ .
... 1'\
.',1
/1
, /
I 0
00
0 U
Te
"..
. 1\
T
V,!"
/1
/L
1'-,
,I
! V
~
l .
~
" V
'1
46
45
44
43
42
41
A
HM
ED
AB
AD
l~
46
1
//1
k/1
1\1
20
0 u
re
~ 4
5
/1
~"'-l',,---
t/
C-L···l.~
J'
~:L
1/1
/-
l--k
...
--., 4
4
'[ I~
ooo~
UT~-~~
~
43
~l
1
I.,A
~t
42
1
~ I
00
2
3 4
6 B
9
10
11
1
2
46~--~~--------------------------~
JO
DH
PU
R
45
1
120
0 U
Te
1"[
42
l I [' ..
... '
. j
l .... '"
I >
'1-,._
/~
\ i
If
ooo~ ~
~~.+
", 1
\T
/
' .. /
f\··l
/ 1,,
\, ~
\ I
... ,
. /
~ \
I -
"'.i
;,
./'
'i
J 1",-,
~"";
44
43
41
I
40
1
(n)
i'
I 2
3 4
5 8
7 8
9 1
0
11
1
2
1 4
4 Ip
AT
IAL
A
/1'"
,.:
.. ..I12
00UT~
L .
I''''.
1 "/
""
1 1
. ,.
....
/ .1
, I
J'"
,
43
42
41
40
39
/ 1
·'
L ·1
,,
! '\
J!
I" ..
' ,
00
00
UT
e '''\
l . \
' 1
1"
,
'.,'
.
'. 1
) /
'
V']
J
48,--
----
----
----
----
----
----
----
----
----
----
---,
BH
OP
AL
...
tl)
43
I ...
'-. 1
.... "
11
2QO
ure
1/
1 i".,
/j
, '-.,1
. ..'
" r,_
1 1/
r ·
.t..
i
t i'.
, .
f //1
5.
"'~"
''''
-..
...
' ..... 1
1/
. )-
00
00
ure
",1
, 1
1 1
/1
1 .... "1
/1
1
'".1
. ¥
I 1
/1
J
42
41
40
39+(~~~~~~~~~~~~~~~
2 3
4 5
6 7
8 e
'0
11
1
2
47
I L
ue
KN
OW
46
12
00
ure
4
5
!//'l
/l
1 .\
. 1
1/
..l' "
'1'.
,.1
/1
l~ ...
. ~-....
.......
,,-'
..
1' ..
..... 1
,)./
V/ J.
0
00
0 u
re
"'.'
\!\
1 ' ...
. ,'i
/ l
/1
J." .
.~/
V
"I
/1
~, _
l/
/ ,~
t,/I
(0)
, ,
, •
• i
44
43
42
41
2
3 4
5 6
7 8
9 1
0
11
1
2
46TI----------------~------------~
SR
,NA
GA
R
T r
.' . 1
" T
v'
l 1
1.12
00 U
Te
l"]'
L
-1
T
l· 1
//1
1"
'."'.1
l' ,
T,/
/"
,', 1
I,
1
1/l
00
00
UT~'I
1 1
l/~
1",
';1
, T
/ /J
'1
. J
1 1
/ "''
'''1
1,/1
J
44
43
~2
41
40
41
I,
'"
"i"
(p)
, i'
i ,
, ,I
)8
I
(q)
I 3e~ ,
, ,
__ ,
,(r)
, .~,
, "
1 2
3 4
5 6
7 8
, 1
0
11
1
2
'i:i .
..
5 ti
1-8'~
1'0
' 1'1
1'
2 1
2 3
4 5
6 7
8 9
10
1
1
12
" J
f· ... '
.
Mo
nth
Of Y
ear
Mon
th o
r Yea
r M
on
th o
r Yea
r F
igu
re
3.3
(j-r
):
Mo
nth
-to
-mo
nth
va
ria
tio
n
of
hD
for
Ko
lka
ta,
Ah
me
da
ba
d,
Bh
op
al,
G
uw
ah
ati
, Jo
dh
pu
r,
Lu
ckn
ow
, D
elh
i,
Pa
tia
la,
an
d S
rin
ag
ar.
co
nti
nu
ou
s lin
es
wit
h s
olid
tri
an
gle
re
pre
sen
t th
e v
ari
ati
on
s fo
r 0
0:0
0 U
TC
an
d t
he
sa
me
wit
h o
pe
n
squ
are
fo
r 1
2:0
0 U
TC
Cl ~
...g 3- w 00
U
-J
- E ~ -..., .c
Cl
.- CU
.c
u .-.., V) .- - ~ t:S ... (0
.c
o ..., <U ;:
u.-
----
----
----
----
----
----
----
--,
TR
IVA
NO
RU
M
20
18
16
14-1
1
20
0U
TC
l A-
... T.
-
,,' Itd
::-i
~/Y'
r!~,
rhl
J V-
--t
10.1
1 0
00
0 U
TC
(a)
8 I
I I
I i
I I
i I
I I
I t
2 3
4 5
6 7
8 9
10
11
12
22
,-----------------------------------,
MA
NG
AL
OR
E
20
18
16
14
-I
' J
. A~f-j;
, '
12 -I
t
12
00
UT
e 1.
-+1
!--:.:.e
:::..:.::t
:--f
1 1
0 -I
r .....
. I--~tP
,/ I
" 1
~r<'::tT
00
00
UT
e
Id)
8~I_.~._~-r,_~~-r~~-._r,_~~,_~~
23
"
6 8
9 1
0
11
12
24
,_------------------------------------,
HY
DE
RA
BA
D
22
18-1
I I)
J)-f'I:~
~"':
I 11
/!
I I
/'j-
-A
16
-I'
i!
: /
' 1"
-,t /'
--
, 't
J/
" "
'.
14
-I 11
l-
'/rO
O U
Te
r
-·t"'--f"~j
1">
1 I
12-1
I .[
(g)
10
~, ri'
I I
j ,
i i
, i
I
20
22
,--------------------------------------,
PO
RT
BL
AIR
20
18
18
T
1 T:l
/Tc~ __
~_l~
---i
-f'-
l--N
1
2.1
i::
::j/l-
VL
I I
lOO
OO
UT
C
14
10
(b)
8 I
I I
I I
I i
I ,
I'
I I
1 2
3 4
5 6
1 8
9 1
0
11
12
20
,----------------------------------------,
CH
EN
NA
I
18
}/L.
t, l
/It j
1,~r
12
-1 t:>
,roo U
TC
If
't""--
.~ .. _
f 't"
'J)'f'
! ,y
!
16
14
10
,/ t 0000
UT
e 8
(e)
641-.~~~~.-~~,_~~,_~~_r~~~
2 4
56
8
9 10
11
1
2
20
'V
ISH
AK
HA
PA
TN
AM
18
16
14-1
1
t--1
2 -I
t 1
/1 "
'V.:J
>'i
1 f\
UOO
UT
;!'
IT I
I {<
Ico-
-1
10
-I
'f>
T
r I
"'I'
!'r"
,·-"
j "
."
00
00
"«
I (h
) 641-,~~_r~~_r~~-r~~~_r,_r,-r~
10
,
BA
NG
AL
OR
E
::~ t r I r
L_
T2
00
UT
C
r 1
4 -I
' -
---
--
J ,"
'V1J--
V!1~-k
l---r-
' r
' %
.00
0 U
Te
L~!
10
.1 1
1------
1 l
l 8
(c)
6 I
I i
i, I'
I I
2 3
4 5
6 7
8 9
10
11
1
2
20
,----------------------------------------,
GO
A
18
16
jf-c
:J."
1 f
t '-~
'~",,-
,;I:,~
. "
f 20
0-;-U
TC
Tr
f 1 ~T __ -
:
t,,-.
i I
/1
t--·-.
.i. ",
"I;
~'>"'.'
.:/
±
! i
tooo
ou~
,
14
12
10
8
(1)
6 I
I i
'i
1 'i
I I
2 3
4 5
6 7
8 9
10
11
1
2
22,-------------------------------------~
20
18
16
14
12
10
8
MU
MB
AI
12
00
UT
C
_ ./
,1
, 1
i.1
J·;~
~!o~
;~:~
I
; ~. __
-,
rl "'
" ,
1 ' f.:
.." I,
j "' 1
±
<: --
-; .
T"!
t-t
• 0
0 (I
' !
: 2
3 ..
S 6
7 8
9 1
0
11
12
2 3
.. 5
6 7
8 9
10
11
1
2
2 J
4 "
6 7
8 9
10
11
1
2
Mo
nth
Of
Yea
r M
onth
Of
Yea
r M
on
th O
f Y
ear
Fig
ure
3.4
(a-i
):
Mo
nth
-to
-mo
nth
va
ria
tio
n o
f hw
fo
r T
riva
nd
rum
, P
ort
bla
ir,
Ba
ng
alo
re,
Ma
ng
alo
re,
Ch
en
na
i, G
oa,
Hyd
era
ba
d,
Vis
ha
kha
pa
tna
m,
an
d M
um
ba
i. co
ntin
uo
us
lines
wit
h s
olid
tri
an
gle
re
pre
sen
t th
e v
ari
ati
on
s fo
r 0
0:0
0 U
TC a
nd
th
e s
am
e w
ith
o
pe
n s
qu
are
fo
r 1
2:0
0 U
TC
n.
;::r
-,
'--6 3- w 00
.j:>
..
22
, K
OL
KA
TA
20
18
- ~
16
-i
r .I:
---J;
"/+
~
12
00
UT
C
IV'
f 1
4 -I
,I,~
-I
/"
/; J~ _
_
"; !+tl:;
V j"1
" _u
le!"
I
B I
I I
I I
I I'
I I
I I
I I
2 J
4 5
6 7
8 9
10
11
1
2
~
22
'O
UW
AH
AT
I
.c
20
C
l ';
18
.c
U
16
t;
1-1
f~I{==I;
:i:~;\'~
.-
14 1
...
,-'
00
00
UT
C
r CJ
'-"
7"
:;:;?
1:;
u 1-,:
1,-1
I , N
~
10
1
' -
'1-'
, ~
-)
.c
-o
8 I
tm)
1 2
J 4
5 6
7 8
9 1
0
11
1
2
," D",H~
". uJ I J,-il/I
-::::f
~~ 1
4'"
/
~
! /'
00
00
UT
C
'-::
, '
~
u.
CJ
3 2
0
18
22
,-----------------_
__
__
__
__
__
__
__
__
__
__
_ --,
AH
ME
DA
BA
D
22
'BH
OP
AL
20
"~,=u.c,
! [ J
Lt
::: 1--l
{~-~
~1
10
'" 0
00
0
Te
1 1
I
I (k
). 'i
• if'
I 8
I i
i I i,
I
2 3
4 5
6 7
B
9 1
0
11
12
20
.. ~ I
I } -f
I ,\'''' u
" 1
6-1
1\ I /j/
i '\/ :\
1
,,~ \/
/ "t
V\:
" I 1
:::
1 j
I I .I
. ... c
'''f'-
j 8~~~~~~(1)~~~~~,2·
2 3
6 7
4 5
8 9
10
1
1
12
22
,-------------------------------------------,
JO
DH
PU
R
22
lLU
CK
NO
W
I" l
20
'"
I ,~
'1\ ri,
l I
'1_,
12
0jT
Ci
/)1 ~,
>,~
,,.
,V(-
'~OO u.c
1, '\1
I f 1/(1
I
\!,,
, r /
' 1
2""
.:'/
1
i l-'o
,J.,,-
I ,/
r
I!
I 1
0 -I
1
i i (~)
i if'.
i •
8 •
I i
i ,
I i
18
16
2 3
4 5
6 7
8 9
10
1
1
12
2
3 4
5 8
7 8
9 1
0
11
1
2
20
,----------------------------------------,
PA
TIA
LA
2
0
, SR
INJlG
IIR
18
1
8"
1 11
20
0 U
TC
'I
A\l
/1 , T
l//r'i~
t ' r
16
14
12
:::
I" Lnu
~H-++l
:!>"J
10
'"
T, /
1.,/1
,J/ 1
00
00
UT
C ~~
" "
I ,/t
8 t-'f
1
I ')-
-,-'--t
(r)
8 I
IP)
I • j
(q)
I &1
I
I I'
I I
I I
I I
I I I
i ,
i i
I ,
i'
i ,
, j
i 'Q
I,'
i ,
i j
i ,
I·
i I
i i
I •
I 2
3 4
6 8
7 8
9 1
0
11
1
2
::~ Hc:j
I H1
,,~ 1
00
00
u'c
! I 8
.J
2 3
• 5
8 7
8 9
10
11
1
2
1 2
3 4
6 a
7 8
9 1
0
11
12
Mo
nth
or Y
ear
Mo
nth
Of
Yea
r M
on
th O
f Y
ear
Fig
ure
3.4
(j-r
): M
on
th-t
o-m
on
th v
ari
ati
on
of
hw f
or
Ko
lka
ta,
Ah
me
da
ba
d,
Bh
op
al,
Gu
wa
ha
ti,
Jod
hp
ur,
Lu
ckn
ow
, D
elh
i, P
atia
la,
and
Sri
na
ga
r. c
on
tin
uo
us
line
s w
ith
so
lid t
ria
ng
le r
ep
rese
nt
the
va
ria
tio
ns
for
00
:00
UTC
an
d t
he
sa
me
wit
h o
pe
n s
qu
are
fo
r 1
2:0
0 U
TC
n.
;::r
--,
I{; 5- v.. 00
V
I
86
3.2.1 Dependence of hD and hw on Surface Temperature
The morning to evening difference in characteristic heights could be attributed mainly
to the corresponding changes in the atmospheric temperature. While this dependence is
distinctly observable for ho• the dependence is rather weak for hw. Then it would be possible
to generate an empirical relation for hD in terms of surface temperature which will be useful
for the application even on a day-to-day basis. Though a systematic morning to evening
difference is not observable for hw. it would be convenient to treat this also analogous to ho
keeping in mind that eq. (3.5) was also proposed on these lines. The only additional input
required in this case would be the value of surface temperature. Once the necessary inputs
for the model viz. surface pressure, water vapor partial pressure and surface temperature are
fed, the parameters Nso. Nsw, ho• and hw can directly be generated using the appropriate
relationships. With this motivation empirical models for ho and hw are developed based on
linear regression. Figure 3.5 shows a scatter plot depicting the temperature dependence of ho
for different RIM stations along with the best fit line obtained through the regression
analysis in the following form
ho = hDO + go Ts (3.8)
The values of hoo and go for all the 18 stations are presented in Table 3.4. The values of hoo
are around 40 km and are fairly constant for all the stations. The values of go, the
temperature gradient of hD is around 0.14 for an these stations.
Following a similar procedure, the empirical relation connecting hw with Ts is
established based on the observed dependencies as depicted in Figure 3.6. As can be seen
from this figure. the scatter of points is relatively large, and the correlation of hw with Ts is
inferior. This could be attributed to the fact that water vapor is not well mixed in the
atmosphere and its altitude variation is not exactly similar to that of the other major gas
constituents in the atmosphere, not withstanding the fact that higher atmospheric
temperature is conducive even for holding more water vapor and extending the profile to
higher altitude. Large-scale advection of water vapor during the monsoon period is another
factor responsible for the observed low correlation between hw and Ts.
Note that, in developing a quartic model for Nw. no theoretical justification was
provided by Hopfield for adopting such a relationship. It was just adapted analogous to that
for hydrostatic component mainly for the convenience in modeling. Keeping these factors
47~--------------------------------, R = 0.99 SO = 0.017 TRIVANORUM
44
43
42
(a) 41+-~--r-----.--r~r-~~--~-.--~~
22 24 26 28 30 32 34
47~--------------------------------,
46
45
44
! 43
'-' .... .= 42 ~
R = 0.999; SO = 0.027 BANGALORE
.. ~ 41+-~~~r-~r-~~(C~I'-~.-~.-~r-~ .= 16 CJ
18 20 22 24 26 26 3D 32 .. ~ 47 .--R-:-0-.9-9-S-0--=-O-.O-24-2-5--------A-HM--EO-A-B-A-O-,
... ~ 46 U ~ ... 45 ~ .= U 44
to 43 Q
42
12 16 20 24 28 32 36 40 44
~.------------------------------, R ~ 0.99 SO = 0.031 DELHI
46
45
44
43
"2
34
47~------------------------------, R = 0.96 SO = 0.067 PORTBLAIR
46
45
44
43
42
(bl 41
22 23 24 25 26 27 28 29 30 31
47 R=0.99; SO = 0.029 KOLKATA
46
45
44
43
42
(d) 41+-~--.-~--.-~--r-----r--r-.~
10 15 20 25 30 3S
47.-----------------------------~ R = 0.999 SO = 0.023 GUWAHATI
46
45
44
43
42
10 12 14 16 16 20 22 l4 26 28 30
45,---------------------------, R = 0.999 SO = 0.063 SRINAGAR
44
43
42
41
40
41+-~-r~-.~-.~(g~J-.~-, __ ~.-~ 39+-~r-~r-~~-(~~~)~~~.-~~~ 10 15 20 25 30 35 40 -6 0 6 10 15 20 25 30 35
Surface Temperature (OC)
87
Figure 3.5: Scatter plots along with best-fit regression line showing the surface temperature dependence of dry characteristic height (hD)
chayter 3 88
17 17 R = 0.21; SO = 0.88 TRIVANDRUM R = 0.36; SO = 1.006 PORTBLAIR
16 16
15 15 .. .. ... 14 r ... .. ; 14 ... ... 13 ... 13 ..
... ... 12 12 • • • • ... .. • 11 11 ... • 10 10
9 (a) (b)
9 22 24 26 28 30 32 22 23 24 25 26 27 28 29 30 31 32
17 17 R = 0.69; SO=1.17 BANGALORE R = 0.43; SO = 1.51 KOL.KATA
16 16 .. , .... 15 • 15
... ~ .", •
14 • 14 • • • •• .. • 13 • 13
'" -~ 12 • • 12 •• '-' 11 • 11 .... • • .r: • 10
... .. ~ 10 .... ... (c) (d) ~ .c 9 9
15 20 25 30 35 40 10 15 20 25 30 35 I;,j .... .... 17 17 III .... R=O.6; so= 1.2 AHMEOABAO R = 0.62; SD c 1.37 GUWAHATI .. • ~ 16 16
"'1 .... I;,j .... ... ~ 15 • 15 ... .. .. ... ~ 14 14 .r: ... U • ..
13 • 13 ... .... ... ... ~ ... ... '" ... ~ 12 12 • ...
• ... • ... 11 ... 11 ....
... 10 10
'" 9
(e) 9
(f)
10 15 20 25 30 35 40 45 10 12 14 16 18 20 22 :1.4 26 28 30
18 16 R= O.BO; SO = 1.09 DELHI R = 0.44. SO = 0.B7 SRINAGAR
17 15 •
16 • • • 14
15 • • 13 ... ... 14 • 12
... • 13 11 ... .. ... • ... ...
• .. ... 10 12 • .. • .. ... • .. •
11 • 9
(g) B
(h)
10 oS 0 5 10 15 20 25 30 35 5 10 15 20 25 30 35 40
Surface Temperature (OC) Figure 3.6: Scatter plots along with best-fit regression line showing the surface temperature dependence of wet characteristic height (hw)
89
Table 3.4: Empirical Relations Representing the Temperature Dependencies for the Dry and Wet Characteristic Height Parameters (hD and hw) for Different Stations
Model for hD Model for hw
Station Name ho = hoo + go Ts hw = hwo + gw Ts
hOG (km) go hwo(km) gw Trivandrum 40.86{0.03) O.l40JO.001) 11.58 (1.75) 0.065 (0.064)
Portblair 40.74 (0.02) 0.146 (0.008) 7.355 (3.46) 0.229 (0.13) Bangalore 40.37 (0.03) 1.14610.001) 7.746 (1.25) 0.230 (0.051) Mangalore 40.46 (0.05) 0.141 (0.002) 14.42 (2.18) 0.063 (0.080)
Chennai 40.28 (0.04) 0.146 (0.001) 7.583 (2.84) 0.180 (0.101) Goa 40.33 (0.11) 0.145 (0.004) 10.23 (3.03) 0.068 (0.111)
Hyderabad 39.36 (0.02) 0.141 (6e-4) 7.792 (1.07) 0.311 (0.039) Vishakhapatnam 39.83 (0.06) 0.151 (0.002) 9.737 (3.42) 0.092 (0.125)
Mumbai 40.23 (0.17) 0.148 (0.006) 8.642 (3.17) 0.152{O.l15) Kolkata 40.28 (0.03) 0.147 (0.001) 9.457 (1.53) 0.131 (0.058)
Ahmedabad 40.36 (0.02) 0.143 (7e-4) 9.923 (1.00) 0.120 (0.035) Bhopal 37.92 (0.03) 0.144 (0.00l) 6.978 (1.15) 0.233 (0.044)
Guwahati 40.34 (0.02) 0.144 (ge-4) 8.730 (1.29) 0.198 (0.050) Jodhpur 39.59 (0.49) 0.121 (0.018) 12.10 (0.94) 0.072 (0.033)
Lucknow 39.75 (0.04) 0.146 (0.001) 7.371 (1.82) 0.244 (0.072) Delhi 40.52 (0.02) 0.144 (8e-4) 9.150 (0.7) 0.178 (0.028)
Patiala 38.41 (0.06) 0.140 (0.002) 8.312 (1.05) 0.076 (0.042)
Srinagar 40.14 (0.02) 0.144 (0.001) 11.50 (0.306) 0.046 (0.020)
into account in order to facilitate the modeling, an empirical relation is established for the
variations of hw with Ts similarly that for hD as,
hw = hwo + gw Ts (3.9)
The coefficients estimated through linear regression for each station relating hw to Ts is
included in Table 3.4. While for the dry characteristic height the value of intercept (hoo) is
fairly constant for all the stations (about 40.5 ± 0.24 km) with a mean slope of 0.144 ± 0.002
km/°C, for wet characteristic height the mean intercept (hwo) is about 9.43 km with a slope
of 0.14 kmf'c. The variability of the intercept and slope in the case of hw for different
stations is large compared to that of hD . However, this is not a major constraint because
even with these limitations the above relationship can provide fairly good prediction of hw.
These models for different stations are referred to as "Site-specific Hopfield model" in
further discussions.
Both the Site-Specific Surface model and Hopfield model estimate the zenith
tropospheric delay in terms of surface meteorological parameters. The essential difference
between these two models is that while the former provides this in terms of surface
90
hydrostatic pressure and water vapor partial pressure, the latter includes the effect of surface
temperature also.
3.3 A Unified Model for the Indian Subcontinent
3.3.1 Unified Surface Model
A close examination of empirical relations derived based on the surface model and
Hopfield model for the individual stations show that the variability of the coefficients
connecting the range error with meteorological parameters is rather small. This prompted
development of a Unified Model (each for surface model and Hopfield model) applicable for
all these stations which hence will be applicable for the entire region encompassing these
stations. Though such unified models could be inferior to the site-specific models, its
potential increases because of the fact that one simple equation will be sufficient for the
entire region. Following the same methodology adopted for individual stations, the relation
between surface meteorological parameters and dry component of range error is sought by
pooling together the data from all the eighteen stations. Figure 3.7a shows the scatter plot of
surface pressure with zenith dry range error, considering all the stations. A simple linear
relation is established based on linear regression (keeping zero intercept) and the slope is
estimated. This yields a simple linear relation termed as the "Unified Surface model" as in
the form
ZHD = (0.00228 ± 2.25xlO-6) x Ps (3.10)
with a correlation coefficient of 0.97. In eq. (3.10) Ps is expressed in hPa and ZHD is
obtained in meters. Following a similar approach, a linear relation connecting ZWD and es
also is established for the whole subcontinent. A scatter plot of ZWD, with the
corresponding values of es is shown in Figure 3.7b. The linear relation of the fonn
(3.11)
is established between the two through regression analysis and evaluated the coefficients Aa
and AI. The values of Aa and Al are, respectively, -0.01978 ± 0.0057 m and 0.01288 ±
2.4x1O-4 m hPa- l. As can be seen from Figure 3.7b, when es is very large (>30 hPa) the
range error estimated through the relation (shown by thin straight line) is consistently lower
than the true range error obtained from ray-tracing (marked points). Similarly a higher
altitude station like Srinagar has lower values of es compared to near sea-level stations
chayter 3 91
(Trivandrum and Portblair). In order to distinguish the values corresponding to Srinagar,
these are plotted in Figure 3.7b with a different symbol (V).
2.6
2.4
E 2.2 '-'
Cl :t N 2.0
1.8
(a) Slope = 0.00228! 1.3915591:-6 Corr. Coeff. = 0.97; er = 0.019
1.6 +-.....--......-,---r-,---"--r--,.....,--.---,.--....--.---l 900 920 940 960 980 1000 1020
Surface Pressure (P sI in mb
U.5
0.4
0.3 E Cl
~ 0.2
0.1
Second Order Polynomial R = 0.95 SO = 0.025 m
Linear Model (b) R = 0.968 SO = 0.027 m
5 10 15 20 25 30 35 ~~ (mb)
0.5..,..-----------------.,..,
0.4
E 0.3
C
~ 0.2
0.1
R = 0.999 SO = 0.0027
(c)
10 20 30 40 50 60 70 80
PW{mm)
Figure 3.7: Regression analysis showing the dependence of dry and wet components of tropospheric zenith range on relevant surface atmospheric parameters considering all the eighteen stations together for generating a unified model: Variation of dry range error with surface pressure (a), Variation of wet range error with surface water vapor partial pressure (b) and variation of wet range error with columnar (precipitable) water vapor (c)
A careful examination of Figure 3.7b shows that the points in this scatter plot are more
aligned to a non-linear increase of ZWD with es. Considering this aspect the applicability of
a second order polynomial is examined to relate ZWD with es, which yielded a relationship
of the form
chaytir 3 92
ZWD= (0.0391±0.013)+ (0.00636±0.OOl) xes +(1.58xl0-4 ±3.75xlO-s)xe/ (3.12)
The nonlinear variation of ZWD by this relationship is shown in Figure 3.7b with the thick
line.
3.3.2 Unified Model in terms ofPrecipitable Water Vapor
As seen in Section 3.1, a linear model relating PW with ZWD will be more appropriate
for predicting ZWD than similar models based on surface values of water vapor partial
pressure, or water vapor density at surface, as far as individual stations are concerned.
However, it should be noted in this context that it will not be possible to get values of PW at
any location whenever prediction of ZWD is required. On the contrary it will be easier to get
values of es or Ps from many places at almost all the time. Hence from the application point
of view such models (relating ZWD with es or Ps) will be more useful even though the error
associated with them will be comparatively larger. One of the ways for estimating PW is by
using a sun-photometer working on the principle of differential attenuation (of water vapor
and other atmospheric species at two optical wavelengths). This instrument operates only
during day time under clear sky conditions and requires regular calibration. Ground based
microwave radiometer, which can work in all weather condition (except heavy precipitation)
though is another potential instrument for measuring PW, is very expensive. The estimation
of PW from radiosonde is not only expensive (recurring expenditure when ever launched)
but also has a poor temporal and spatial resolution (considering the time taken for each
sounding and the balloon drift). The satellite based microwave radiometers though are very
useful over the oceanic regions they will not be available quite frequently. A GPS network
could be a good source of PW especially over the land (not too expensive). This can be
operated in all weather conditions with a very good temporal resolution and does not require
any calibration. In this case, a Unified Model to estimate ZWD from PW will be useful.
Figure 3.7c shows a scatter plot of PW with true zenith wet delay (estimated from
refractivity profiles derived from radiosonde) clubbing all the eighteen stations. A linear
model was developed based on the regression analysis, which yielded a relationship of the
form
ZWD = (0.0064 ± 4.4xl 0-6) ~ PW (3.13)
chayter 3 93
with correlation coefficient close to unity. As can be seen directly fonn Figure 3.7c, this
model is significantly superior to the other three models. The mean absolute difference for
different stations is presented in the last column of Table 3.7 along with the corresponding
standard deviation. A similar model was also used by Bevis et al. [1992] for deriving ZWD
from ground based GPS measurements. In his global model, Bevis et al. [1992] evolved a
value of 0.00666 as the ratio of ZWDfPW which varies as much as 20% with location,
season and weather condition [Be vis et al., 1994].
3.3.3 Unified Hopfield Model (Characteristic Height Parameters)
Following a very similar approach a unified empirical model is developed for zenith
tropospheric delay based on the Hopfield method. This essentially involves deriving a
unified model for dry and wet characteristic height parameters in tenns of Ts by grouping
the data from all the eighteen stations. The temperature dependence of hD and hw are
established incorporating the data from all the 18 stations by linear regression which is of
the form
hD ;::: (40.209 ± 0.045) + (0.154 ± 0.002) >; Ts
hw = (10.474 ± 0.29) + (0.111 ± 0.01) >; Ts
(3.14a)
(3.14b)
A typical scatter plot of hD (in km) with Ts (in QC) and hw (in km) and Ts (in 0c)
obtained by combining all the stations is presented in Figure 3.8a and Figure 3.8b along with
the best fitted line represented by eq. 3.14a and eg. 3.14b, respectively. The correlation
coefficient for hD and hw are 0.98 and 0.56, respectively. Comparing the values of these
regression coefficients with those presented in Table 3.4 (site-specific surface model), it can
be seen that the present value of slope is slightly larger than the corresponding values in
Table 3.4 while the value of the intercept is slightly small.
3.4 Validation of Surface Models
3.4.1 Validation of Site-Specific Models by Comparing with Ray-Traced Values
The site-specific models are validated by comparing the model predictions with the
true range errors for the respective station obtained through ray tracing employing the
altitude profiles of refractivity. Sufficient care is taken to ensure that the daily
94
meteorological data have water vapor profile extending up to or beyond la km altitude, and
the temperature and pressure profiles extending up to at least 25 km. The comparison is
quantified by estimating the absolute difference between true and model predicted zenith
delays, a summary of which is presented in Table 3.5. The data for RIM stations are
highlighted with bold face. The mean of the absolute differences are presented along with
the standard deviations. While the mean value indicates the average deviation of the model,
the standard deviation indicates the amount of variability. The accuracy of surface model
based on Ps is comparable to that of the Hopfield model in predicting ZHD. The mean
deviation is lowest for Trivandrum and highest for Jodhpur.
48 18
R = 0.987; SD = 0.187 R=O.S6 SO'" 1.2
46 I~
44 14 - --2 :: ::. ::.
l!: 1> ~ 42 .::: 12 1>
1> 1>
1>
40 JO 1> 1> 11:
(a) (b) 38 8
10 20 3D 40 SO 0 IU 10 30 4U
T/C) T~(C)
50
Figure 3.8: Regression analysis showing the dependence of dry (hD) and wet (hw) characteristic heights on surface temperature CTs) considering all the eighteen stations together
The deviation of Site-Specific models, in case of ZWD, based on es and Ts is about 5-
6 cm for most of the stations. For a few stations like Vishakhapatnam, Mumbai, Jodhpur
and Lucknow the deviation is > 6 cm. Srinagar (being high altitude station) shows
minimum mean absolute difference about -2 cm for ZWD. Among all the RIM stations
considered, Kolkata shows maximum deviation of -6 cm. The best estimate of wet delay
comes from model based on Precipitable Water vapor. The mean absolute difference for
this model is -0.3 cm except at Patiala and Lucknow. By combining the surface models
based on Ps and the wet range error model based on PW, the best estimate of ZTD (with an
error <6 cm) can be achieved for almost all the stations. As an alternative (when PW is not
c'hayter 3 95
available) the surface models based on es or Hopfield model based on es and Ts (which are
more easily available) could be considered, even though the involved accuracies are rather
inferior. Note that, the accuracy achieved by these models also meets basic requirement of
satellite based navigation system discussed in Chapter 1.
Table 3.5: Mean Absolute Difference and the standard deviation of Site-specific Models (fl"om true zenith range error) for Different Stations (RTM stations in Bold), for the Hydrostatic and Non-hydrostatic Components
Hydrostatic delay (ZHD) Non-hydrostatic delay (ZWD)
Station Mean abs. Diff (cm) Mean abs. Diff (cm)
Surface mode Hopfield Surface model Hopfield Surface model based on Ps model based on es model based on PW
Trivandrum 0.17 (0.1) 0.15 (0.1) 5.2 (3.7) 5.1 (3.6) 0.17 (0.1) Portblair 0.36 (0.3) 0.49 (0.4) 4.5 (3.5) 4.6 (3.6) 0.16 (0.1) Bangalore 0.45 (0.2) 0.2 (0.2) 4.0 (3.0) 4.2 (3.0) 0.16 (0.1)
Mangalore 2.3 (1.9) 2.5 (2.5) 4.9 (3.6) 6.3 (4.3) 0.64 (0.3)
Chennai 4.5 (2.7) 4.4 (2.7) 6.1 (4.5) 5.9 (4.4) 0.6 (0.3)
Goa 2.4 (0.5) 1.5 (0.5) 5.4 (3.5) 5.7 (3.9) 0.5 (0.23)
Hyderabad 3.7 (2.0) 3.5 (2.0) 6.7 (4.6) 6.1 (4.4) 0.4 (0.26)
Vishakhapatnam 3.5 (2.6) 3.6 (2.7) 7.6(6.3) 7.5 (6.2) 0.3 (0.2)
Mumbai 3.18 (2.8) 3.12 (2.7) 6.9 (4.3) 7.5 (4.5) 0.51 (0.27)
Kolkata 0.35 (0.2) 0.32 (0.3) 6.0 (4.2) 6.4 (4.3) 0.21 (0.1)
Ahmedabad 0.34 (0.3) 0.33 (0.2) 5.2 (4.0) 5.3 (4.0) 0.24 (0.2)
Bhopal 1.25 (1.2) 1.35 (1.25) 6.5(5.1) 5.8 (4.6) 0.66 (0.3)
Guwahati 0.24 (0.2) 0.25 (0.2) 4.4 (3.3) 5.0 (3.6) 0.20 (0.1)
Jodhpur 11.2 (6.2) 10.0 (5.5) 9.5 (8.1) 9.4(8.1) 0.4 (0.3)
Lucknow 5.7 (5.7) 5.7 (5.8) 8.8 (12.9) 8.6 (12.7) 1.1 (1.3)
Delhi 0.27 (0.2) 0.27 (0.2) 4.0 (3.0) 4.5 (3.6) 0.20 (0.2)
Patiala 2.7 (2.3) 3.4 (2.6) 4.1 (3.0) 5.6 (3.8) 2.0 (0.4)
Srinagar 1.80 (1.7) . 1.09 (1.8) 2.0 (2.0) 1.9 (1.6) 0.19 (0.1)
3.4.2 Validation of Unified Models using Ray-Tracing
The accuracy of unified model is also examined by comparing the estimated delays
with those obtained by ray tracing as well as with other global models based on surface
parameters [Saastamoinen, 1972; Hopfield, 1971; Ifadis, 1986; Mendes and Langley, 1998].
The mean of the absolute difference of the deviations for each station is presented in Table
3.6. These values are larger than those for site-specific surface models presented in Table
3.5, especially for Trivandrum and Ahmedabad. For the case of ZWD, the comparison of
chayter 3 96
Table 3.6: Mean Absolute Difference and the standard deviation of Unified Model and Global Model from True Zenith Range Error for Different Stations, for the Hydrostatic Component
Hydrostatic delay
Station Mean abs. Diff (cm)
Code SAAS Surface Model Hopfie1d
(1972)
Ps Ps model
Trivandrum 2.66 (0.2) 1.01 (0.2) 1.24 (0.27) Portblair 2.57 (0.4) 0.94 (0.4) 1.10 (0.40) Bangalore 1.03 (0.3) 0.45 (0.2) 0.30 (0.20) Mangalore 2.33 (2.13) 2.29 (1.88) 2.87 (2.66)
Chennai 4.09 (2.5) 3.84 (2.35) 5.15 (2.83) Goa 2.10 (0.47) 2.40 (0.47) 1.00 (0.43)
Hyderabad 2.28 (2.45) 2.23 (2.25) 3.08 (2.75) Vishakhapatnam 4.70 (2.80) 4.33 (2.75) 5.90J2.84)
Mumbai 2.86 (2.50) 2.79 (2.26) 3.39 (3.15) Kolkata 0.84 (0.4) 0.97 (0.4) 0.80 (0.4)
Ahmedabad 0.70 (0.4) 1.10 (0.4) 0.99 (0.5) Bhopal 12.10 0.73) 11.70 (1.73) 13.25 (1.7 1)
Guwahati 0.85 (0.6) 0.79 (0.5) 0.93 (0.8) ]odhpur 16.02 (6.68) 15.66 (6.66) 17.25 (6.83)
Lucknow 7.40 (6.26) 7.08 (6.19) 8.45 (6.45) Delhi 1.90 (0.3) 0.28 (0.2) 0.47 (0.25) Patiala 7.76 (3.11) 7.44 (3.03) 8.81 (3.40)
Srina2ar 2.40 (2.6) 3.20 (2.0) 2.90 (2.0) All stations
2.90 (3.60) 2.90 (3.4) 2.98 (4.15) together
both linear (eq. 3.11) and second order (eq. 3.12) models are presented in Table 3.7. In case
of ZWD the deviation of unified model is comparable to that of Site-Specific model. Among
different models suggested for ZWD, the second order polynomial Unified model yields a
better prediction. Considering all the eighteen stations together the mean of the absolute
difference for SAAS and Unified model from the true value is - 2.9 cm. In the case of wet
delay, the Unified models developed in the present analysis are much superior to the
currently available models [Ifadis, 1986 and Mendes and Langley, 1998]. The mean absolute
difference of the Unified model based on es varies from about 4 cm to 9 cm whereas that in
case of other two global models is in the range 5 cm to 15 cm. This study indicates the
region specific models (for tropics) are superior to the available global models developed
97
Table 3.7: Mean Absolute Difference and the standard deviation of Unified Models and Global Models from True Zenith Range Error for Different Stations, for the Nonhydrostatic Component
Non-hydrostatic delay Mean abs. Diff (cm)
Mendes Ifadis Surface 2na order
Hopfield Surface
(1997) (1986) Model- Model-
Model Model-
es es PW
Trivandrum 8.15 9.18 5.1 4.8 4.98 0.17 (4.7) (4.8) (3.6) (3.7) (3.5) (0.17)
Portblair 7.51 8.49 4.5 4.4 4.40 0.16 (5.2) (5.3) (3.5) (3.5) (3.3) (0.13)
BangaIore 8.12 8.08 4.3 4.9 4.20 0.57 (3.8) (3.8) (2.9) (2.8) (2.8) (0.21)
MangaJore 7.97 9.08 4.2 4.12 5.25 0.37
(4.62) (4.82) (3.4) (3.27) (3.93) (0.23)
Chennai 8.70 9.98 5.4 5.20 4.35 0.26
(5.14) (5.40) (4.0) (3.88) (3.25) (0.20)
Goa 7.27 8.18 4.9 4.76 4.86 0.22
(4.84) (5.46) (3.0) (2.86) (2.67) (0.17)
Hyderabad 15.53 16.24 11.3 10.92 11.11 0.52 (5.90) (5.91) (5.6) (5.53) (5.55) (0.29)
Vishakhapatnam 8.96 9.81 7.2 6.92 6.90 0.26
(6.98) (7.41 ) (5.6) (5.47) (5.44) (0.20)
Mumbai 10.81 12.07 6.0 5.45 6.20 0.18 (5.38) (5.77) (3.9) (3.83) (3.99) (0.13)
Kolkata 11.21 12.44 5.9 5.06 5.70 5.70 (5.9) (6.2) (4.1) (4.0) (4.0) (4.0)
Ahmedabad 7.62 8.58 4.8 4.8 5.00 5.00 (5.6) (5.8) (3.7) (3.9) (3.9) (3.9)
Bhopal 7.54 8.07 6.18 6.14 6.03 6.03
(6.71) (6.91) (4.5) (4.47) (4.58) (4.58)
Guwahati 11.87 12.92 6.1 4.3 5.90 5.90 (5.7) (5.9) (4.1) (3.7) (4.0) (4.0)
Jodhpur 14.32 15.51 9.8 9.73 9.97 9.97 (8.45) (8.63) (8.3) (8.50) (8.06) (8.06)
Lucknow 14.08 15.26 9.28 9.27 9.54 9.54
(13.33) (13.53) (14.2) (13.08) (13.00) (13.00)
Delhi 9.18 9.96 5.0 4.0 4.90 4.90 (6.7) (6.9) (3.9) (3.1) (3.9) (3.9)
Patiala 4.43 5.13 7.35 5.88 5.23 5.23
(3.29) (3.50) (4.7) (4.06) (3.75) (3.75)
Srinagar 5.49 4.78 2.0 2.1 1.97 1.97 (3.8) (3.8) (1.9) (1.8) (1.8) (1.8)
All stations 9.89 10.85 4.75 5.81 6.10 6.10 together (6.64) (6.86) (3.74) (5.43) (5.33) (5.33)
chayter 3 98
mainly based on the meteorological data from mid and high latitude regions. All the models
for ZHD show poor performance for the stations Bhopal, Jodhpur, Lucknow and Patiala
could be due to poor radiosonde data quality at these stations. Excluding these, the mean
absolute difference for ZHD estimated using Surface model (- 2.1 cm) is less than the other
two models (- 2.5 cm).
3.5 Variability in Zenith Tropospheric Delay and its Deviation from
Model Derived Values for Different Stations
The dry and wet component of the tropospheric zenith delay shows a rather systematic
annual variation, though the amplitude and phase show a variation with latitude. It would be
interesting in this context to examine the distribution of these delays on a day-to-day basis
for a period of one year to assess the extent of variability. Values of ZHD, ZWD and ZTD
(estimated through ray tracing) on different days are grouped at small intervals and the
percentage of occurrence of each group is estimated. Similarly, in order to examine the
model accuracies, the deviation of model predicted values of ZHD, ZWD and ZTD from the
true values estimated as above are also estimated for different days in a year and the
frequency distribution of these deviations are also examined. Figure 3.8 shows the
distribution of true ZHD obtained through ray-tracing and dZHD (true value of ZHD - ZHD
derived from Unified model). Figure 3.8a and Figure 3.8b shows the plots of these
distributions for different stations. Similarly Figure 3.9a and Figure 3.9b shows the
distribution of ZWD and dZWD for these stations, while Figure 3.lOa and Figure 3.l0b
shows the respective distributions of ZTD and dZTD. The distribution of ZHD is more-or
less similar for all these stations except for Srinagar (for which it is truncated in the upper
half). The distribution is slightly skewed in its lower half for Bangalore while it is skewed in
its upper half for Bhopal. The distribution of model deviation for the ZHD is generally sharp
and symmetric for Ahmedabad and station south of it. For station north of it, the distribution
peaks at the highest value. The distribution of ZWD is more-or-Iess symmetric for
Bangalore, Portblair and Trivandrum. For other stations the distribution is rather flat
indicating that all values are equally probable. For stations like Mangalore,
Vishakhapatnarn, Mumbai and Delhi, the distribution shows a double hump feature. The
distribution of the deviation of model from true value of ZWD is more or less symmetric for
w !I! ;;! ~
u is .. w
U ! ~ u is .. ~ ~ ..
50
1 TR
IVA
NO
RU
M
-I TR
IVA
ND
RU
M
" P
OR
TB
LA
IR
50
'PO
RT
BL.
-to
IF!:
50
1 B
AN
OA
LO
RE
5
0
BA
NG
AL
OR
E
" ,,'
I' I
"I
I "
-"
,. ,.
"
n~ "
" ,.
n
" " ..
.. • 1
"I Ut
11 Ill.
, ",
23
2
23
3
23
4
1:5
2.0
, '"
" ,.
J ~"
l MA
NG
AL
OR
E
50
O
OA
" I
GOA
" "
" "
" "
.. ..
, ,
.. ..
, '"
no
'"
'"
HY
DE
RA
BA
D
" "
VlS
HA
KH
AP
AT
NA
M
MU
MB
AI
IMU
MB
AI
HY
DE
RA
BA
D
" "
" "
" ,.
,. "
~ R.
,. " 2
24
2
2.
23
2
, . .. ~
.,
• ..
ZH
D (
cm
) 6
ZH
D !
cm
) Z
HD
!cm
) ,\
ZH
D !
cm
) Z
HD
(c
m)
.... Z
HD
(cm
)
" ~
~ '}.
W
;
Figu
re 3
.8a
: F
requ
ency
dis
trib
utio
n o
f tr
ue
Zen
ith H
ydro
sta
tic D
elay
(Z
HD
) e
stim
ate
d t
hro
ug
h r
ay-
tra
cin
g a
long
wit
h t
he
dis
trib
utio
n
\0
of
the
m
odel
d
evi
atio
ns
fro
m
the
tr
ue
va
lue
for
Trl
van
dru
m,
Por
tbla
lr,
Ban
galo
re,
Man
galo
re,
Che
nnal
, G
oa,
Hyd
era
bad,
\0
Vls
hakh
apat
nam
, an
d M
umba
l
50
'K
OL
KA
TA
.. w
~ 3
0
~
o ,.
g ,. "
i ~ ,.
GU
WA
HA
TI
150
, D
EL
HI
w"
o I ,.
o g ,.
,.
..
23
2
23
4 " !
GU
WA
HA
Tl
" .. DE
LH
I
o!
f.
L U
JI
I ,
o,Jlo-~~"",
:Z20
2
24
2
2&
-'"
-~
-'>
" 50
, AHMEOA~O
AH
ME
DA
BA
O
.. ..
" ,.
,. ,.
" "
01.1
;1
."
1'1[1
-0
1 _
IIIH
UU
"jI
22
4
2Z1
U&
2
30
2
32
-3
·2
·1
0
50
'JO
DH
PU
R
" JO
OH
PU
R
.. ..
" "
,. ,. "
_ _
OIlH"1CII'~
n 2
00
2
10
2
20
2
30
·2
5·2
0·1
5-1
0 -a
0
11 1
0
50
, P
AT
IAL
A
50
'PA
T.A
LA
.. " ,. "
.. " ,. " OI!IIIII~'
21
0
21
1
22
2
22
1
_12
...
-4
0 ..
" B
HO
PA
L
.. " ,. " 01
PI
IH
II,
20
0
20
4
20
8
SR
INA
GA
R
601 B
HO
PA
L,
" .. " "
-"
-"
50
'S
RIN
AG
AR
.. " ,. "
~
DUll
U
IIIW
UrJ
·1
2
...
-4
0
ZH
D (
cm
) IJ,
ZH
D (
cm
l Z
HD
(cm
) 11
. ZH
D (
cm
) Z
HD
(cm
) 4
ZH
D (
cm
l
Figu
re 3
.8b:
Fre
quen
cy d
istr
ibu
tio
n o
f tr
ue
Ze
nith
Hyd
rost
atic
De
lay
(ZH
D)
est
ima
ted
th
rou
gh
ra
y-tr
aci
ng
alo
ng
wit
h t
he
d
istr
ibu
tio
n o
f th
e m
od
el
de
via
tion
s fr
om
th
e t
rue
va
lue
fo
r K
olka
ta,
Ah
me
da
ba
d,
Bho
pal,
Gu
wa
ha
tl, l
od
hp
ur,
lu
ckn
ow
, D
elhi
, P
atla
la,
and
Sri
naga
r
~
l..g ~
w - 8
w ~ g ... w
u I .. I ...
" " "
TR
IVA
NO
RtJ
M
501 P
OR
TB
LA
IR
" " " " , "
" C
HE
NN
AI
CH
EN
NA
I
" "
" "
" "
. "
" "
50
G
OA
5
01
GO
A
" "
" "
" Bn~
~~ "
" "
01
.'
."h
UlIlI
oll'1
UI!_
1
0 2
0
30
.0
11
0 -1
2
-"
0 6
12
" M
UM
BA
I
" " "
" M
UM
BA
I
" " " " oI
OIE
!!"!!
']U
·1
1·1
2"
0 •
12
1.
ZW
O (
cm
) 6.
%W
O (
cm
) %
WO
(cm
) ,\
ZW
O (
cm
) Z
WO
(cm
) 6.
%W
O I
cm
)
Fig
ure
3.9
a:
Fre
quen
cy d
istr
ibu
tio
n o
f tr
ue
Ze
nith
We
t D
elay
(Z
WD
) e
stim
ate
d t
hro
ug
h r
ay-
tra
cin
g a
lon
g w
ith
th
e d
istr
ibu
tion
of
the
m
od
el
de
via
tio
ns
fro
m
the
tr
ue
va
lue
fo
r T
riva
nd
rum
, P
ort
bla
lr,
Ban
galo
re,
Man
galo
re,
Che
nnai
, G
oa,
Hyd
era
ba
d,
Vis
ha
kha
pa
tna
m,
an
d M
umba
i
n ~
I~ 'S-~ - o -
90,
50
, AH
ME
OA
BA
D
" "I"H
O""
~
BH
OP
AL
'}
.
"I "I
" ..
1'1
w
1/ "I
-
"I
" "
" 1'1
" ~ ~
"I
1111
1 "
I .1
1 "
I "
I ...
"I
I.
I I
g "
.. '"'
....
'"'
.1
1 ...
'"'
••• 1
111
':1 ull
lll~1l
':1
Ill.! Ill
lln ':
." ."
,
" "
" "
30
4O
5
0
60
_2
0 _i
D ,
" "
SO,
GU
WA
HA
TI
50lG
UW
AH
AT
I 5
0 JO
OH
PU
R
" "
70
ILU
CK
NO
Vl/.
JO
mlP
UR
L
UC
KN
OW
" "
" "
~
" ~ "
.I~I "
J~n
" "
" ~
" ~
_IU
u "
" "
" "
,. g ..
" "
" "
, ,
, 40
-40-
20 0
2
04
06
0
" "
" "
" ·H
.,
to 2
0 3
0 4
0 5
0 8
0 7
0 8
0
DE
LH
I 5
01 S
RIN
AG
AR
w
4U
( .. ,
:rA
TIA~
• 5
0r
AT
1AL
A
" ..
50
] S
RIN
AG
AR
" "
u ~ ~ u g .. "I
"I
"I
11.
"I
11
"I
" "I
"I
Ill
n "
ullL
" ••
" " ,
., ,
, 1
0
20
]0
'I
D
50
6
0
·H
ZW
D (
cm
t A
ZW
D (
cm
) Z
WO
(cm
) A
%W
O (
cm
) Z
WD
Icm
) t>
ZW
O t
em)
Figu
re
3.9b
: F
requ
ency
d
istr
ibu
tio
n
of
tru
e
Ze
nith
W
et
Del
ay
(ZW
O)
est
ima
ted
th
rou
gh
ra
y·tr
acl
ng
a
lon
g
wit
h
the I <
5 d
istr
ibu
tion
of
the
mod
el d
evia
tions
fro
m t
he t
rue
val
ue f
or
Kol
kata
, A
hmed
abad
, B
hopa
l, G
uwah
atl,
Jodh
pur,
lu
ckno
w,
N
Del
hi,
Pat
iala
, an
d S
rln
ag
ar
50
I TfU
VA
ND
RU
M
.. I '" ~ J g '"
HV
De
RA
BA
O
.. ~
~ ,.
~ .. " o 1r·
1II.
".
~~n
, ..
n ,.
" T
RIV
AN
DR
UM
5O
'PO
RT
BL
AIR
.. .. , "
Ill."
"
HV
OE
RA
BA
O
V1
SH
AK
HA
PA
TN
AM
"
501B
AN
GA
LO
FlE
50
tA
NG
AL
OR
E
.. 1
.. " " " .11
11.11
111 !J
,,.
, .. "
. 50
, G
OA
" 1
DOA
.. ..
"
.Hnt "
" "
" "
01
..
"'-
'"
11'
1'1
O.
_ .....
. 2.
40
25
0'
2.tI0
2
70
2
80
·1
5 -
10
-.5
0 5
10
15
" M
UM
BA
I
.. ..
" " "
_10
0 1
0
2.0
ZT
O (
cm
) /\
ZT
O (
cm
) Z
TO
(cm
) ...
ZT
O (
cm
) Z
TO
(cm
) ...
ZT
O (
cm
)
FIgu
re
3.10
a:
Fre
quen
cy d
istr
ibu
tio
n o
f tr
ue
Ze
nith
Tro
posp
heri
c D
elay
(Z
TD
) e
stim
ate
d t
hro
ug
h r
ay-
tra
cin
g a
lon
g w
ith
the
d
istr
ibu
tio
n
of
the
m
od
el
de
via
tion
s fr
om
th
e t
rue
val
ue
for
Trl
van
dru
m,
Po
rtb
lalr
, B
anga
lore
, M
anga
lore
, C
henn
al,
Goa
, H
vder
abad
. V
ish
akh
ao
atn
am
. an
d M
umba
i
19,
'-S ::-- ~ - 8
" " ~ "
~ 2
0
" " (I
I 1
][]I
)I1
! iL
l! :1
Il!
"tn
'A
n ,,
.n "
'.n ,
1Nl
" " ~ "
~ 2
0
" "
GU
WA
HA
TI
OD
lll
lm
C1
UI!
!I
25
0
26
0
27
0
21
0
...
50
, D
EL
HI
w
" u ~
" B
g
,.
" " •
50
, D
EL
HI
" " ,. "
50,
AH
ME
DA
BA
D
" " ,. " -1
2
-e 0
• U
80
'JO
DH
PU
R
"" JO
DH
PU
R
" "
" "
,. ,.
" "
" B
HO
PA
L
" " " " 01
FI.,
I, I
J U
2
11
2
21
2
40
2
52
"' lU
CK
NO
W
" " ,. ..
50
, B
HO
PA
L
" " ,. " -2
" ·1
15
.a
0 I
eo
,LU
CK
NO
W
" " " " O
Ul'I
r'I
l!II
Irlln
nn
o
lrla
1l,
r!2
O
"ULIP
UI
';
10
1_
.... !n
_
21
0 2
25
2
40
25
5 2
70
21
5 ~O
-15
(I
U
30
2
40
2
52
2
84
2
76
2
11
..e
O -
eO 4
0 ·
20
0
20
4
0
SO,
PA
TlA
LA
5
0'P
AT
IAL
A
" "
.. "
" ,.
" .. "
50,
SR
INA
GA
R
" " ,. ..
." "
~n
" 0
1 _,
,11 •
• "111
__
ol~
' !
" ~,
I t
ill
olrl
La
1211
1'1
" n
24
0
25
2
21
4
21
6
2&1
_20
_1
0
(I
10
2
0
23
0
24
0
25
0
26
0
o 11·
... P
4I
." -I
(I
2
00
2
10
~~..
...
ZT
D (
.. m
) ....
ZT
O (
cm
) Z
TO
(cm
) A
ZT
D (
cm
) Z
TD
(cm
) (\
ZT
D I
cm)
g,.,
l..g - :)- w
Figu
re
3.10
b:
Fre
quen
cy d
istr
ibu
tio
n o
f tr
ue
Ze
nith
Tro
posp
heri
c D
elay
(Z
TO
) e
stim
ate
d t
hro
ug
h
ray-
tra
cin
g
alon
g w
ith
th
e
dis
trib
utio
n o
f th
e m
odel
de
via
tion
s fr
om
th
e t
rue
val
ue f
or
Kol
kata
, A
hmed
abad
, B
hopa
l, G
uw
ah
atl,
Jo
dh
pu
r, L
uckn
ow,
Del
hi, I ~
Pat
lala
, an
d S
rin
ag
ar
105
Srinagar, Delhi, Guwahati, Ahmedabad and Kolkata. For other stations the distribution is
rather broad or non-symmetric. The distribution of ZTD is fairly symmetric for Srinagar,
Ahmedabad, Bangalore, Portblair and Trivandrum while Delhi; it also shows a double hump
like feature. For other stations it is rather skewed in the upper half. The model deviation of
ZTD is fairly systematic for all stations and is more governed by the shape of the ~WD
frequency distribution.
3.6 Summary
The neutral atmosphere introduces a delay in the propagation of microwaves, which
leads to an error for ranging in GPS based aircraft navigation, significant mainly during
landing and take-off. The neutral atmosphere being non-dispersive at GPS signal
frequencies, estimation of tropospheric delay is possible only through modeling. Simple
linear relationships are established for ZHD in terms of surface pressure and for ZWD in
terms of three humidity parameters; the surface water vapor pressure, surface water vapor
density and precipitable water vapor based on data from eighteen locations over the Indian
subcontinent as well as that from an Island station in the Bay of Bengal representative of
different climatic zones. In addition to these linear models based on meteorological data the
"characteristic height" parameters are modeled for different stations to implement the
Hopfield model. Using this model, the prediction of tropospheric error could be
accomplished for any given altitude above the surface. Taking note of the relatively small
variability of the coefficients in the above Site-Specific models, applicability of a Unified
model for the entire region encompassing these stations is evolved by pooling the data from
different stations. A ftrst order model in tenns of surface pressure for ZHD and a second
order nonlinear model for ZWD in tenns of water vapor partial pressure at the surface are
found to be most suitable for the prediction purpose. Following a similar approach unified
models are also developed for the Hopfield characteristic height parameters. All these
models are validated by comparing the model predictions with the true range errors obtained
through ray tracing the refractivity profiles derived from radiosonde data. As far as Site
Specific models are concerned minimum deviation for the model prediction of ZHD is
observed at Trivandrum (- 0.17 cm) while the maximum deviation is observed for Srinagar
(- 2 cm). The validity of the Unified Surface model is also examined by comparing modeled
delay values with true ray-traced values as well as with other global models. In the case of
chayt(f 3 106
ZWD the deviation (- 5.81 ± 5.43 cm) of Unified Surface model is comparable to that of
site-specific model. The Unified surface model is found to be superior to the global models
when applied to Indian region.
****